Internet of vehicles coverage hole identification method and system, electronic device, and storage medium

Abstract

The present application relates to the technical field of communication network, and especially relates to a kind of Internet of Vehicles coverage hole identification method, system, electronic equipment and storage medium.The method comprises: obtaining the multi-source heterogeneous data in the Internet of Vehicles environment;According to the multi-source heterogeneous data, the network experience of each target vehicle in the target time point in the Internet of Vehicles coverage range is predicted, and the network experience expected data of each target vehicle is obtained;The actual data of the network experience of each target vehicle at the target time point is obtained;For each target vehicle, the network experience expected data of the target vehicle at the target time point and the actual data of the network experience of the target vehicle are compared to determine whether the position of the target vehicle is a coverage anomaly point according to the comparison result obtained;All coverage anomaly points in the same space-time range are aggregated to obtain the Internet of Vehicles coverage hole.The present application realizes the active, real-time and fine identification of communication hole, so as to provide accurate basis for network optimization.

Description

Technical Field

[0001] This invention relates to the field of communication network technology, and in particular to a method, system, electronic device, and storage medium for identifying coverage gaps in vehicle networks. Background Technology

[0002] With the rapid iteration of intelligent transportation and vehicle-to-everything (V2X) technologies, the vehicle's role as a mobile communication terminal is becoming increasingly prominent. Real-time data interaction between vehicles, between vehicles and roads, and between vehicles and the cloud has become fundamental to realizing core functions such as autonomous driving, traffic scheduling optimization, driving safety warnings, and intelligent cockpit experiences. As the core carrier of V2X data transmission, the coverage quality of operators' 4G and 5G wireless communication networks directly determines the reliability, timeliness, and integrity of data transmission. However, the highly dynamic and heterogeneous deployment environment of V2X places higher demands on the quality of wireless network coverage. High-speed vehicle movement leads to frequent switching of wireless channels. Signal obstruction or multipath interference exists in places such as underground parking lots and tunnels. The irregularity of road network topology, the density of base station deployment, and the uneven distribution of vehicles can easily create wireless signal coverage gaps, or communication holes, within specific time and space. Communication holes directly lead to interruptions in data transmission between vehicles, a surge in latency, or data packet loss. This not only causes inaccurate autonomous driving decisions, delayed traffic information updates, and a decline in vehicle user experience, but may even lead to serious driving safety accidents, becoming a core bottleneck restricting the large-scale application of V2X.

[0003] Currently, the analysis and identification of wireless network coverage problems mainly rely on the following methods: road tests or fixed-point tests by optimization engineers. This method uses professional equipment to measure along specific routes in the field, which can obtain relatively accurate coverage data. However, it is costly, has limited coverage, is difficult to sustain, and cannot achieve real-time monitoring of large-scale road networks. Network-side alarms and counter statistics are based on macro-level analysis of base station operation indicators. However, this method can only reflect overall problems at the network or cell level, is not sensitive to coverage degradation in localized small areas, and suffers from detection lag. User complaint-driven methods rely on users actively reporting network experience problems. However, by the time a problem occurs, it has already affected the user experience, making it a passive response maintenance method that cannot proactively provide early warnings. Based on the perception analysis of human network users, this method uses measurement report data from ordinary mobile phone users to assess network coverage. However, mobile terminals and vehicle communication modules have different sensitivity to signals, and their business logic is different. Therefore, human network data analysis has only limited reference value and is not entirely applicable to vehicle network scenarios. In summary, existing technologies lack an effective means to proactively, in real-time, and with fine detail discover dynamically changing weak or blind spots in network coverage in vehicle network environments. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method, system, electronic device and storage medium for identifying gaps in vehicle network coverage, so as to solve the above-mentioned technical problem.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: a method for identifying coverage holes in a vehicle-to-everything (V2X) network, comprising: acquiring multi-source heterogeneous data in a V2X environment; predicting the network experience of each target vehicle within the V2X coverage area at a target time point based on the multi-source heterogeneous data to obtain expected network experience data for each target vehicle; acquiring actual network experience data for each target vehicle at the target time point; for each target vehicle, comparing the expected network experience data and the actual network experience data of the target vehicle at the target time point to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison result; and aggregating all coverage anomalies within the same spatiotemporal range to obtain a V2X coverage hole.

[0006] The beneficial effects of this invention are as follows: By acquiring multi-source heterogeneous data in the vehicle-to-everything (V2X) environment, this invention predicts the network experience of the target vehicle at the target time point and compares it with the actual experience. Based on the comparison results, it actively determines coverage anomalies and then aggregates and generates coverage holes. This achieves proactive, real-time, and refined identification of communication holes, overcoming the shortcomings of existing technologies that rely on road tests, alarms, or complaints, such as passive lag, high costs, and inability to locate local micro-areas. It can promptly discover hidden holes, providing accurate basis for network optimization, thereby effectively improving the reliability of V2X communication and the safety of autonomous driving.

[0007] Based on the above technical solution, the present invention can be further improved as follows.

[0008] Furthermore, the step of predicting the network experience of each target vehicle within the coverage area of ​​the vehicle network at a target time point based on the multi-source heterogeneous data to obtain the expected network experience data of each target vehicle includes: performing spatiotemporal interpolation completion processing on the multi-source heterogeneous data to obtain new multi-source heterogeneous data; constructing features based on the new multi-source heterogeneous data to generate multi-dimensional fusion features corresponding to each target vehicle; and for each target vehicle, predicting the network experience of the target vehicle at the target time point based on the multi-dimensional fusion features corresponding to the target vehicle to obtain the expected network experience data of the target vehicle.

[0009] Furthermore, the spatiotemporal interpolation completion processing of the multi-source heterogeneous data to obtain new multi-source heterogeneous data includes: for each target vehicle in the multi-source heterogeneous data where wireless network measurement data is missing, constructing a spatial association graph of the target vehicle with the target vehicle as the target node and the base stations and roadside units around the target vehicle as surrounding nodes; for each target vehicle in the multi-source heterogeneous data where wireless network measurement data is missing, performing data prediction using a pre-trained graph neural network based on an attention mechanism according to the spatial association graph of the target vehicle to obtain the wireless network measurement data completion value of the target vehicle; and filling in the multi-source heterogeneous data based on all the wireless network measurement data completion values ​​to obtain new multi-source heterogeneous data.

[0010] Furthermore, the multi-dimensional fusion features include at least a signal stability index, an environmental interference coefficient, and a vehicle-road cooperative quality score; wherein, the signal stability index is data that quantifies the correlation between signal fluctuations and vehicle movement, the environmental interference coefficient is data that quantifies the impact of environmental factors on signal quality, and the vehicle-road cooperative quality score is data that quantifies the quality of the vehicle-road communication link.

[0011] Furthermore, for each target vehicle, the network experience of the target vehicle at a target time point is predicted based on the multi-dimensional fusion features corresponding to the target vehicle to obtain the expected network experience data of the target vehicle. This includes: inputting the multi-dimensional fusion features corresponding to the target vehicle into a pre-trained time-series prediction model to generate a first expected network experience value; constructing a dynamic road network topology map using the target vehicle and surrounding vehicles, base stations, and roadside units as nodes; propagating spatial information through a pre-trained dynamic graph neural network based on the dynamic road network topology map and the first expected network experience value to generate a correction vector; correcting the first expected network experience value based on the correction vector to obtain a second expected network experience value; and obtaining the expected network experience data of the target vehicle based on the first expected network experience value and the second expected network experience value.

[0012] Furthermore, for each target vehicle, the expected network experience data and the actual network experience data of the target vehicle at the target time point are compared to determine whether the location of the target vehicle is a coverage anomaly based on the comparison result. This includes: determining whether the location of the target vehicle is an anomaly based on the multi-dimensional fusion features corresponding to the target vehicle using a pre-trained binary classification model to obtain a first judgment result; calculating a first deviation between the first expected network experience value and the actual network experience data of the target vehicle, and calculating a second deviation between the second expected network experience value and the actual network experience data of the target vehicle; comparing the first deviation with a deviation threshold to obtain a second judgment result; comparing the second deviation with the deviation threshold to obtain a third judgment result; and making a collaborative decision based on the first judgment result, the second judgment result, and the third judgment result through a voting mechanism to determine whether the location of the target vehicle is a coverage anomaly.

[0013] Furthermore, the deviation threshold is obtained as follows: the vehicle network coverage area is divided into multiple spatiotemporal partitions based on time and space dimensions; historical operation data within each spatiotemporal partition is acquired; for each spatiotemporal partition, the upper limit of the prediction error at a given confidence level is predicted using quantile regression based on the historical operation data within the spatiotemporal partition, which serves as the initial threshold for the spatiotemporal partition; the initial threshold corresponding to the spatiotemporal partition to which the target vehicle currently belongs is corrected based on the real-time environmental data in the multi-source heterogeneous data to obtain the deviation threshold.

[0014] Furthermore, the aggregation of all coverage anomalies within the same spatiotemporal range to obtain vehicle network coverage holes includes: acquiring multiple coverage anomalies, each containing a geographic location and a timestamp; dynamically adjusting the clustering radius based on the road type where the coverage anomalies are located; using an improved DBSCAN algorithm to spatially cluster the coverage anomalies to generate candidate hole regions; and using a streaming clustering algorithm to perform real-time streaming aggregation and merging of the candidate hole regions according to a preset time window and spatial grid to generate vehicle network coverage holes and their geographic location boundary information.

[0015] Furthermore, before aggregating all coverage anomalies within the same spatiotemporal range to obtain the vehicle network coverage gap, the method further includes: for the same geographical location, if coverage anomalies from multiple vehicles of different models continuously exist within a consecutive preset number of time windows, then the geographical location is determined to have passed multi-vehicle verification; for geographical locations that have passed multi-vehicle verification, the wireless network measurement data reported by the vehicle's infotainment system, the communication quality data reported by the roadside unit and the corresponding vehicle's infotainment system, and the regional signal quality data detected by the roadside unit are obtained; when the wireless network measurement data reported by the vehicle's infotainment system indicates that the signal strength is lower than a first preset threshold, and the communication quality data indicates communication anomalies or the regional signal quality data indicates that the signal strength is lower than a second preset threshold, the coverage anomaly point corresponding to the geographical location is confirmed as a coverage anomaly point for aggregation.

[0016] Furthermore, the method also includes: acquiring the vehicle lighting status of each target vehicle; when the meteorological data in the multi-source heterogeneous data shows that there is target weather, and the vehicle lighting status of more than a preset proportion of target vehicles in the same area is in fog light on or low beam on, environmental attenuation compensation is performed on the network experience expectation data based on the meteorological data in the multi-source heterogeneous data.

[0017] Furthermore, the step of predicting the network experience of each target vehicle within the coverage area of ​​the vehicle network at a target time point based on the multi-source heterogeneous data to obtain the expected network experience data of each target vehicle also includes: when the target vehicle is in a tunnel, predicting the network experience of the target vehicle based on the vehicle location information in the multi-source heterogeneous data and a pre-constructed signal attenuation reference curve to obtain the expected network experience data of the target vehicle; wherein, the pre-constructed signal attenuation reference curve is a curve constructed based on historical signal measurement data of the tunnel, used to reflect the attenuation trend of the reference signal received power with the tunnel depth.

[0018] Furthermore, the method also includes: acquiring electromagnetic interference quantification indicators reported by multiple target vehicles, and generating an interference source spatial heat map based on the electromagnetic interference quantification indicators; performing spatial clustering analysis based on the interference source spatial heat map to identify local coverage holes.

[0019] To address the aforementioned technical problems, the present invention also provides a vehicle network coverage hole identification system, comprising: The data acquisition module is used to acquire multi-source heterogeneous data in the vehicle-to-everything (V2X) environment; The prediction module is used to predict the network experience of each target vehicle within the coverage area of ​​the Internet of Vehicles at a target time point based on the multi-source heterogeneous data, and obtain the expected network experience data of each target vehicle. The data acquisition module is used to acquire the actual network experience data of each target vehicle at the target time point; An anomaly detection module is used to compare the expected network experience data of the target vehicle and the actual network experience data of the target vehicle at the target time point for each target vehicle, so as to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison result. The hole determination module is used to aggregate all coverage anomalies within the same spatiotemporal range to obtain the coverage holes of the vehicle network.

[0020] To address the aforementioned technical problems, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the vehicle network coverage hole identification method as described above.

[0021] To address the aforementioned technical problems, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the vehicle network coverage hole identification method as described above. Attached Figure Description

[0022] Figure 1 This is a flowchart of the vehicle network coverage hole identification method of the present invention; Figure 2 This is a schematic diagram of the vehicle network coverage hole identification system of the present invention; Figure 3 This is a schematic diagram of the electronic device of the present invention. Detailed Implementation

[0023] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0024] As mentioned earlier, the continuity and stability of wireless communication coverage are prerequisites for the normal operation of the Internet of Vehicles (IoV). Accurately identifying and locating communication gaps is of vital practical significance and application value for optimizing wireless communication network deployment, improving the reliability of IoV communication, and ensuring the safe and efficient operation of intelligent transportation systems.

[0025] However, existing technologies lack effective means to proactively, in real-time, and with precision discover dynamically changing weak or blind spots in network coverage within the connected vehicle environment, particularly addressing deep-seated, hidden issues where signal strength is acceptable but service experience is poor. Numerous gray areas affecting autonomous driving safety and user experience remain unidentified and unresolved in a timely manner.

[0026] Example 1 Based on this, such as Figure 1As shown, this embodiment provides a method for identifying gaps in vehicle network coverage, including: S101. Acquire multi-source heterogeneous data in the vehicle networking environment.

[0027] S102. Based on multi-source heterogeneous data, predict the network experience of each target vehicle within the coverage area of ​​the vehicle network at the target time point to obtain the expected network experience data of each target vehicle.

[0028] S103. Obtain the actual network experience data of each target vehicle at the target time point.

[0029] S104. For each target vehicle, the expected network experience data of the target vehicle at the target time point will be compared with the actual network experience data of the target vehicle, so as to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison results.

[0030] S105. Aggregate all coverage anomalies within the same spatiotemporal range to obtain vehicle network coverage gaps.

[0031] The core idea of ​​this method is that network coverage holes are essentially caused by a significant deviation between a vehicle's actual network experience at a specific geographical location and its expected experience based on history and context. By acquiring multi-source heterogeneous data in a vehicle-to-everything (V2X) environment, the method predicts the network experience of a target vehicle at a target time point and compares it with the actual experience. Based on the comparison results, it proactively identifies coverage anomalies and aggregates them to generate coverage holes. This achieves proactive, real-time, and refined identification of communication holes, overcoming the shortcomings of existing technologies that rely on road tests, alarms, or complaints, such as passive lag, high costs, and inability to locate small local areas. It can promptly discover hidden holes, providing accurate basis for network optimization, thereby effectively improving the reliability of V2X communication and the safety of autonomous driving.

[0032] In this embodiment, the multi-source heterogeneous data mainly includes vehicle-side multi-sensor data, roadside unit (RSU) data, meteorological and electromagnetic environment data, signaling acquisition big data, and wireless network measurement data. Wireless network measurement data includes at least one of Measurement Report (MR) data and Call Detail Record (CDR) data.

[0033] Specifically, vehicle-side multi-sensor data originates from the onboard T-Box / ADAS system. Key metrics include real-time vehicle speed, acceleration, steering angle, altitude, and headlight status, such as low beam headlights being on in a tunnel. This data supplements the dynamic attributes of the spatial context, such as changes in signal obstruction angles during high-speed turns, helping to distinguish between signal fluctuations caused by vehicle movement and actual coverage gaps. Roadside equipment data originates from 5G-V2X roadside units. Key metrics include communication latency between the RSU and the vehicle-mounted system, packet loss rate, RSU's own load, and signal transmission power. This data provides redundant vehicle-to-road communication data in areas without base station coverage (such as the middle of a tunnel), addressing the data loss issue when base station signals are interrupted. Meteorological and electromagnetic environment data are derived from meteorological monitoring stations / operator wireless network signal electromagnetic interference monitoring indicators. Core indicators include rainfall, wind speed, visibility, and surrounding electromagnetic interference intensity, such as electromagnetic radiation along high-speed rail lines. These are used to quantify the impact of environmental factors on signal quality. For example, heavy rain can cause a 3-5 dB drop in the Signal to Interference plus Noise Ratio (SINR), preventing environmental interference from being misjudged as coverage gaps. Signaling big data is derived from Deep Packet Inspection (DPI) data collection. Core indicators include service uptime and service continuity for vehicles within the network. This data provides service uptime and continuity data for vehicles under various conditions and geographical locations as a basis for judging network coverage. Wireless network measurement data (MR / CDR) is derived from wireless measurement reports. Core indicators include the overall network coverage at the vehicle's geographical location. All data is organized by time slices (e.g., every 10 seconds) and correlated with the vehicle's geographical location.

[0034] Optionally, in this embodiment, based on multi-source heterogeneous data, the network experience of each target vehicle within the coverage area of ​​the vehicle network at a target time point is predicted to obtain the expected network experience data of each target vehicle. This includes: performing spatiotemporal interpolation completion processing on the multi-source heterogeneous data to obtain new multi-source heterogeneous data; constructing features based on the new multi-source heterogeneous data to generate multi-dimensional fusion features corresponding to each target vehicle; and for each target vehicle, predicting the network experience of the target vehicle at the target time point based on the multi-dimensional fusion features corresponding to the target vehicle to obtain the expected network experience data of the target vehicle.

[0035] Optionally, in the embodiments, spatiotemporal interpolation completion processing is performed on the multi-source heterogeneous data to obtain new multi-source heterogeneous data, including: for each target vehicle in the multi-source heterogeneous data where wireless network measurement data is missing, constructing a spatial association map of the target vehicle with the target vehicle as the target node and the base stations and roadside units around the target vehicle as surrounding nodes; for each target vehicle in the multi-source heterogeneous data where wireless network measurement data is missing, based on the spatial association map of the target vehicle, performing data prediction through a pre-trained graph neural network based on an attention mechanism to obtain the wireless network measurement data completion value of the target vehicle; and filling in the multi-source heterogeneous data based on all the wireless network measurement data completion values ​​to obtain new multi-source heterogeneous data.

[0036] To address the lack of MR data in areas such as tunnels and under elevated roads, an improved spatiotemporal interpolation completion model is proposed. This model is an improvement upon the standard graph neural network. Specifically, instead of using a static graph, a spatial correlation graph is dynamically constructed every 5 seconds based on the real-time location and relative speed of the vehicle, base station, and RSU, combined with event-driven updates (e.g., triggering local graph reconstruction when the vehicle enters a new base station cell, rather than a full real-time update, minimizing computational overhead). The nodes in the graph include the three base stations and two RSUs surrounding the target vehicle. The initial edge weights are determined by Euclidean distance, and a speed attenuation factor is introduced (e.g., when the vehicle moves away from the base station at a speed greater than 60 km / h, the edge weight is reduced by an additional 20%) to more accurately reflect signal change trends under high-speed movement.

[0037] A GATv2 (Graph Attention Network v2) architecture is adopted to replace the traditional GAT, achieving dynamic and more expressive attention weight allocation. The model uses the spatiotemporal features of the target node (missing data point) as the query vector, and the features of surrounding nodes as keys and values ​​to calculate attention coefficients. Finally, weighted aggregation is performed to obtain the predicted values ​​of the missing Reference Signal Received Power (RSRP) and Reference Signal Received Quality (RSRQ). The algorithm is trained end-to-end using historical data, and the loss function is the mean squared error between the predicted and actual values, ultimately achieving a data missing rate of less than 3%.

[0038] Optionally, in the embodiments, the multi-dimensional fusion features include at least a signal stability index, an environmental interference coefficient, and a vehicle-road cooperative quality score; wherein, the signal stability index is data that quantifies the correlation between signal fluctuations and vehicle movement, the environmental interference coefficient is data that quantifies the impact of environmental factors on signal quality, and the vehicle-road cooperative quality score is data that quantifies the quality of the vehicle-road communication link.

[0039] By modally fusing vehicle-side sensor data with network metrics, three types of highly discriminative derived features are generated. Among them, the signal stability index = (current RSRP - standard deviation of RSRP over the past 5 seconds) / vehicle speed is used to quantify the correlation between signal fluctuations and vehicle movement. For example, if the signal fluctuation is large but does not decrease significantly at high speeds, holes can be excluded.

[0040] The environmental interference coefficient is calculated as: Measured SINR / Meteorological interference baseline. The meteorological interference baseline is a dynamically adjusted reference value, obtained as follows: First, a lightweight regression model (such as XGBoost) is trained based on historical data (including meteorological information and corresponding SINR values). The model takes real-time meteorological data such as rainfall, visibility, and wind speed as input and outputs the expected SINR baseline value under those meteorological conditions. The environmental interference coefficient is the ratio of the measured SINR to this baseline value, used to quantify the real-time impact of environmental factors on signal quality.

[0041] The vehicle-to-infrastructure (V2I) cooperative quality score is used to evaluate the quality of the vehicle-to-infrastructure (V2I) communication link, and mainly includes two specific indicators. The specific calculation formula is: V2I cooperative quality score = (RSU communication delay / base station communication delay) × RSU load factor.

[0042] Among them, RSU communication latency and base station communication latency: These two latency data are directly obtained from the communication module logs of the vehicle terminal. When the vehicle terminal communicates with the base station and RSU, it records the round-trip latency of each signaling interaction. By parsing these logs in real time, the communication latency with the RSU and base station is extracted.

[0043] RSU Load Factor: This factor is obtained from the RSU device's management information database or the northbound interface of the operation management platform, and reflects the current resource utilization of the RSU (such as the number of connected users, CPU utilization, etc.).

[0044] This score combines link latency and RSU load, and is used to help determine whether communication quality is affected by voids in areas with no base station coverage or weak coverage.

[0045] Finally, the three types of derived features mentioned above, along with other features extracted from multi-source heterogeneous data, are spliced ​​together to form multi-dimensional fusion features.

[0046] Optionally, in this embodiment, for each target vehicle, the network experience of the target vehicle at a target time point is predicted based on the multi-dimensional fusion features corresponding to the target vehicle to obtain the expected network experience data of the target vehicle. This includes: inputting the multi-dimensional fusion features corresponding to the target vehicle into a pre-trained time-series prediction model to generate a first expected network experience value; constructing a dynamic road network topology map using the target vehicle and surrounding vehicles, base stations, and roadside units as nodes; propagating spatial information through a pre-trained dynamic graph neural network based on the dynamic road network topology map and the first expected network experience value to generate a correction vector; correcting the first expected network experience value based on the correction vector to obtain a second expected network experience value; and obtaining the expected network experience data of the target vehicle based on the first and second expected network experience values.

[0047] The time-series prediction model employs an encoder-decoder Transformer architecture, with input consisting of 64-dimensional multidimensional fused features from the past 30 time slices (each 10 seconds). In this embodiment, these 64-dimensional features comprise the following three parts: Core metrics (20 dimensions): including raw metrics extracted from MR and DPI data such as RSRP, RSRQ, SINR, uplink and downlink throughput, Transmission Control Protocol (TCP) retransmission rate, and service latency; Derived features (30 dimensions): including the aforementioned signal stability index, environmental interference coefficient, vehicle-road cooperative quality score, and one-hot coding of vehicle speed, acceleration, steering angle, and road type (such as highway, tunnel, etc.); Environmental and contextual features (14 dimensions): including real-time rainfall, wind speed, visibility, time (hour, day of the week), and embedding vectors of geographic grid IDs.

[0048] The model outputs predicted network perception metrics for the next 1-3 time slices (10-30s), which are the expected values ​​for the first network experience. Specifically, these include the communication quality index (calculated based on RSRP and SINR weighted) and the service experience index (calculated based on throughput and latency weighted).

[0049] In this embodiment, the Transformer encoder-decoder structure is not directly applied but rather customized and improved. Specifically, a temporal attention mask matrix is ​​introduced into the Transformer's self-attention calculation. This matrix is ​​generated based on external events (such as tunnel entrance / exit points and overpass ramps in map data), forcing the model to assign higher attention weights to features from specific past time slices (such as 5 seconds before entering the tunnel) when predicting the output of time slices near these event points. This solves the problem of standard Transformer's prediction lag in abrupt scene changes.

[0050] Furthermore, the encoder layer is designed with a dual-branch structure. The short-time branch takes the data from the most recent 5 time slices as input and uses a 1D convolution with a kernel size of 3 to capture short-term fluctuations in the signal. The long-time branch takes the data from all 30 time slices as input and uses a standard Transformer layer to capture the long-term trend of the signal. The outputs of the two branches are concatenated and fused at the end of the encoder before being fed into the decoder. This structure can simultaneously capture both the instantaneous jitter and periodic patterns of the signal.

[0051] This model uses historical data for supervised learning. The training dataset consists of 30 consecutive time slices as input samples, followed by 1-3 time slices as labels. The loss function is a combination of the mean absolute error (MAE) between the predicted and actual values ​​and a smoothed L1 loss. End-to-end training is performed through backpropagation and the Adam optimizer, ultimately reducing the prediction error by 35%-40% compared to the traditional ST-GNN model.

[0052] In the second stage, a dynamic road network topology map is constructed based on the prediction results from the first stage. Nodes include vehicles, base stations, and RSUs (Roadside Units). Edge weights are updated in real-time according to vehicle location (e.g., the edge weight decreases by 15% for every 100m increase in distance between a vehicle and a base station). A three-layer Graph Convolutional Network (GCN) is used to correct the base predictions. The specific correction process is as follows: The communication quality index and service experience index of each vehicle predicted in the first stage are used as the initial features of the nodes. These initial features are input into a three-layer GCN (Global Networking Community). Each layer of the GCN updates its own node features by aggregating the features of neighboring nodes (surrounding vehicles, base stations, RSUs) based on the current dynamic graph structure. This process simulates the spatial mutual influence of network coverage quality. For example, if a vehicle has a poor signal, there is a high probability that surrounding vehicles will also have poor signals.

[0053] The final output of GCN is a correction vector with the same dimension as the initial features. The base prediction value from the first stage is added to this correction vector to obtain the final corrected prediction value, which is the expected value experienced by the second network. This residual connection design allows the model to focus on learning the spatial correlation error between the base prediction value and the true value, rather than learning predictions from scratch, thereby improving learning efficiency and stability.

[0054] By incorporating road topology (such as the upper and lower levels of elevated roads and the direction of tunnels) into the graph structure as prior knowledge, the model avoids misjudging the inability of vehicles on the lower level of the elevated road to receive signals from the base station on the upper level as a void. Instead, clustering and void confirmation are performed only on the reported abnormal spatiotemporal points, significantly reducing the amount of computation.

[0055] The graph structure is updated every 5 seconds, consistent with the data acquisition and feature calculation cycle. At a vehicle speed of 120 km / h, the vehicle's position changes approximately 166 meters within 5 seconds, sufficient to capture the topological changes caused by high-speed movement. Real-world testing shows that at this update frequency, the prediction accuracy can be improved by 15%-20% after correction.

[0056] Optionally, in an embodiment, for each target vehicle, the expected network experience data and the actual network experience data of the target vehicle at the target time point are compared to determine whether the location of the target vehicle is a coverage anomaly based on the comparison result. This includes: determining whether there is an anomaly at the location of the target vehicle based on the multi-dimensional fusion features corresponding to the target vehicle using a pre-trained binary classification model to obtain a first judgment result; calculating a first deviation between the first expected network experience value and the actual network experience data of the target vehicle, and calculating a second deviation between the second expected network experience value and the actual network experience data of the target vehicle; comparing the first deviation with a deviation threshold to obtain a second judgment result; comparing the second deviation with the deviation threshold to obtain a third judgment result; and making a collaborative decision through a voting mechanism based on the first judgment result, the second judgment result, and the third judgment result to determine whether the location of the target vehicle is a coverage anomaly.

[0057] In this embodiment, a lightweight XGBoost binary classification model is introduced as a third decision-maker. This model takes real-time features (such as real-time RSRP, SINR, vehicle speed, weather, etc.) as input and directly outputs a judgment result (0 or 1) indicating whether it is abnormal.

[0058] The first deviation between the expected network experience value and the actual network experience data of the target vehicle is calculated, and the second deviation between the expected network experience value and the actual network experience data of the target vehicle is also calculated. The first deviation is compared to a deviation threshold; if the first deviation is greater than the deviation threshold, a second judgment result of anomaly is obtained. The second deviation is then compared to the deviation threshold; if the second deviation is greater than the deviation threshold, a third judgment result of anomaly is obtained. Based on the first, second, and third judgment results, a collaborative decision is made through a voting mechanism to determine whether the location of the target vehicle is a coverage anomaly point. For example, a coverage anomaly point is only determined when at least two judgment results are anomaly, avoiding false alarms caused by data noise from a single model.

[0059] To reduce computational overhead and improve processing efficiency, the following method can be used for judgment: First, XGBoost is used for rapid screening to identify potential outliers. If XGBoost determines it to be normal, the point is directly considered normal, and Transformer and DGNN are not called again to reduce computational overhead. For the screened potential outliers, Transformer+DGNN joint prediction is performed to obtain accurate predicted values. For the Transformer and Dynamic Graph Neural Network (DGNN) models, the difference between the predicted value and the actual measured value is calculated. When the difference exceeds a dynamic threshold, the model votes for an outlier. A three-vote voting mechanism is used; if the number of outlier votes reaches a preset threshold, it is ultimately determined to be a covered outlier.

[0060] Furthermore, through SHAP (SHapley Additive exPlanations) value analysis, the contribution of each feature to the anomaly detection is output (e.g., RSRP measured value is 3dB lower than the predicted value, contribution is 60%; electromagnetic interference coefficient exceeds the standard, contribution is 30%), providing clear problem attribution for subsequent network optimization (e.g., holes are caused by weak signals, not interference).

[0061] Optionally, in the embodiments, the deviation threshold is obtained in the following way: the vehicle network coverage area is divided into multiple spatiotemporal partitions based on the time and space dimensions; historical operation data in each spatiotemporal partition is obtained; for each spatiotemporal partition, the upper limit of the prediction error at a given confidence level is predicted by quantile regression based on the historical operation data in the spatiotemporal partition, which is used as the initial threshold of the spatiotemporal partition; the initial threshold corresponding to the spatiotemporal partition to which the target vehicle currently belongs is corrected based on the real-time environmental data in the multi-source heterogeneous data to obtain the deviation threshold.

[0062] Specifically, the time dimension is divided into peak hours, off-peak hours, and late-night hours, and the spatial dimension is divided into urban roads, highways, and tunnels / elevated roads. A separate dynamic threshold calculation model is constructed for each spatiotemporal partition, which is a statistical model based on quantile regression. For each spatiotemporal partition, historical data under normal conditions is collected. This normal data is obtained through historical analysis, after removing known malfunctions, complaints, and severe weather events, and represents the normal fluctuation range of the partition under healthy conditions.

[0063] On the historical normal data of this partition, a quantile regression model is trained, with time as the independent variable, to predict the upper limit of the prediction error at a given confidence level (e.g., 95%). This upper limit is the basic threshold for this partition. The basic threshold is based on quantile regression of historical data and is updated daily. Furthermore, a dynamic correction factor is introduced to incorporate real-time environmental features (such as rainfall and cell load) to linearly or nonlinearly adjust the basic threshold, achieving minute-level adaptation.

[0064] For example, during peak hours in the highway zone, the calculated baseline dynamic threshold is: This is twice the standard deviation of normal fluctuations. Considering the increased signal fluctuations caused by multiple vehicles operating concurrently during peak hours, a dynamic adjustment factor is set to relax the threshold for practical applications to... This adjustment factor is set based on statistical analysis of false alarm rates during historical peak periods.

[0065] The threshold is updated every 10 minutes based on the latest 500 normal data points (without anomaly markers) to adapt to dynamic network changes (such as the overall signal decline caused by temporary power reduction of base stations), with a target false negative rate of less than 3%.

[0066] Optionally, in this embodiment, aggregating all coverage anomalies within the same spatiotemporal range to obtain vehicle network coverage holes includes: acquiring multiple coverage anomalies, each containing a geographic location and a timestamp; dynamically adjusting the clustering radius based on the road type where the coverage anomalies are located; using an improved DBSCAN algorithm to spatially cluster the coverage anomalies to generate candidate hole regions; and using a streaming clustering algorithm to perform real-time streaming aggregation and merging of the candidate hole regions according to a preset time window and spatial grid to generate vehicle network coverage holes and their geographic location boundary information.

[0067] Specifically, the improved DBSCAN clustering algorithm replaces the traditional fixed-radius scheme with a dynamic radius. It adjusts the radius based on road type, for example, 20m for urban roads and 50m for highways. This prevents adjacent lanes from being abnormally merged into a single void on urban roads, improving positioning accuracy from the 50m level to the 10m level, achieving sub-lane-level positioning. Furthermore, it introduces a density-adaptive radius, dynamically adjusting the radius based on vehicle density within the clustering area, reducing the radius when density is high, and incorporating road semantic information for refined segmentation, such as lane lines and medians.

[0068] The FStream (Fast Stream Clustering) algorithm is used to process abnormal events in real time by combining a time window (10s) with a spatial grid (10m×10m). Clustering results are output every 10s, and the aggregation delay is within 10s. This allows for the timely detection of sudden gaps, such as signal interruptions caused by sudden base station failures.

[0069] Optionally, in the embodiments, before aggregating all coverage anomalies within the same spatiotemporal range to obtain the vehicle network coverage hole, the method further includes: for the same geographical location, if coverage anomalies from multiple vehicles of different models continuously exist within a consecutive preset number of time windows, the geographical location is determined to have passed multi-vehicle verification; for geographical locations that have passed multi-vehicle verification, the wireless network measurement data reported by the vehicle's infotainment system, the communication quality data reported by the roadside unit and the corresponding vehicle's infotainment system, and the regional signal quality data detected by the roadside unit are obtained; when the wireless network measurement data reported by the vehicle's infotainment system indicates that the signal strength is lower than a first preset threshold, and the communication quality data indicates communication anomalies or the regional signal quality data indicates that the signal strength is lower than a second preset threshold, the coverage anomaly corresponding to the geographical location is confirmed as a coverage anomaly for aggregation.

[0070] In this embodiment, multi-vehicle verification requires that there be at least 3 coverage anomalies of different vehicle types in the same geographical location, and the duration of the anomalies must be at least 2 time windows to eliminate instantaneous interference. By requiring different vehicle types, such as including fuel vehicles and electric vehicles, it avoids group false alarms caused by the failure of a certain brand of vehicle module.

[0071] Fault sources are identified by comparing vehicle-mounted MR (Mean Moving Average) data with roadside unit (RSU) data, enabling cross-validation of equipment. Specifically, MR data (such as RSRP value) reported by the vehicle-mounted unit and communication quality data with the vehicle-mounted unit reported by the RSU are acquired simultaneously, such as Received Signal Strength Indicator (RSSI) and Signal-to-Noise Ratio (SNR) for vehicle-to-infrastructure communication. If the RSRP reported by the vehicle-mounted unit is below a threshold (e.g., -110dBm), the signal is considered weak. Conversely, if the RSSI reported by the RSU for communication with the vehicle-mounted unit is above a threshold (e.g., -80dBm), the RSU communication is considered normal.

[0072] When both of the above conditions are met simultaneously, it indicates that there may be a problem with the vehicle's infotainment system itself (such as antenna failure or decreased module sensitivity), rather than a gap in network coverage at that location. Only when the vehicle's infotainment system reports a weak signal, and the communication between the RSU and the vehicle's infotainment system is also abnormal (such as RSSI being below a threshold), or the RSU also detects overall weak signal in the area, will the location be finally confirmed as a coverage anomaly.

[0073] Optionally, in the embodiments, the method further includes: acquiring the vehicle lighting status of each target vehicle; when the meteorological data in the multi-source heterogeneous data shows that there is target weather, and the vehicle lighting status of more than a preset proportion of target vehicles in the same area is fog lights on or low beam lights on, environmental attenuation compensation is performed on the network experience expectation data based on the meteorological data in the multi-source heterogeneous data.

[0074] When real-time rainfall reaches the heavy rain threshold (e.g., ≥25 mm / h), an empirical attenuation compensation step is triggered. This step, based on historical rainfall attenuation data (the average decrease in SINR under heavy rain conditions), artificially lowers the future SINR prediction output by the Transformer model by 15%-20% before output. This ensures that when the actual SINR decreases due to rainfall attenuation, it remains close to the corrected prediction value, avoiding misjudgment.

[0075] The system acquires the lighting status (fog lights, low beam headlights) uploaded from the vehicle's CAN bus and cross-compares it with meteorological data. If the meteorological data indicates heavy rain and more than 80% of vehicles in the same area have their fog lights / low beam headlights on, it is confirmed as a genuine heavy rain scenario, triggering the compensation logic. If the weather forecast indicates no rain but a large number of vehicles have their fog lights on, it may be a sensor false alarm, and compensation will not be triggered. This verification provides greater certainty about the current environmental conditions, thus more reliably triggering the environmental degradation compensation logic and further reducing false alarms.

[0076] Optionally, in this embodiment, based on multi-source heterogeneous data, the network experience of each target vehicle within the coverage area of ​​the vehicle network at a target time point is predicted to obtain expected network experience data for each target vehicle. The method further includes: when a target vehicle is in a tunnel, based on vehicle location information in the multi-source heterogeneous data and a pre-constructed signal attenuation reference curve, the network experience of the target vehicle is predicted to obtain expected network experience data for the target vehicle. The pre-constructed signal attenuation reference curve is a curve constructed based on historical signal measurement data of the tunnel, used to reflect the attenuation trend of the reference signal received power as the tunnel depth increases.

[0077] In this embodiment, the tunnel is divided into three sections: the entrance section (0-50m), the middle section (50-200m), and the exit section (200m-exit). Different data sources and judgment logic are used for different sections. For the middle section, since the base station signal coverage is basically nonexistent, hole identification mainly relies on RSU data (such as vehicle-to-road communication latency and packet loss rate). For the entrance / exit sections, the base station signal begins to appear or disappear, and the signal fluctuation is severe. At this time, the data from both the base station and RSU are fused, and a weighted voting method is used to jointly determine whether it is a hole. For example, a hole is only identified when the base station signal strength (RSRP) is below a threshold and the RSU communication quality (such as signal-to-noise ratio) is also below a threshold, in order to avoid misjudging instantaneous fluctuations during signal switching as holes.

[0078] For each tunnel, a personalized signal attenuation baseline curve is trained using historically accumulated normal signal data—MR data from periods with no user complaints or fault records. This curve is a regression model that fits the normal attenuation trend of RSRP as distance changes, with the tunnel entrance as the starting point and the tunnel depth as the independent variable. For example, the trained model might show that the RSRP decreases by an average of 8.2 dB for every 100 m of tunnel depth. When the real-time measured RSRP value is consistently (for more than two time windows) lower than 10% of the predicted value of the baseline curve, a coverage hole is identified at that location.

[0079] Optionally, in an embodiment, the method further includes: acquiring electromagnetic interference quantification indicators reported by multiple target vehicles, and generating an interference source spatial heat map based on the electromagnetic interference quantification indicators; performing spatial clustering analysis based on the interference source spatial heat map to identify local coverage holes.

[0080] In this embodiment, electromagnetic interference coefficients reported by multiple vehicles in a spatial grid are collected in real time and mapped onto a geographic grid to form a dynamic spatial heat map of interference sources. By performing spatial cluster analysis (such as DBSCAN) on this heat map, the types of interference can be distinguished.

[0081] Specifically, if the clustered area is huge (e.g., covering several base station cells) and the interference coefficient is generally high, it is determined to be a global interference, which may be caused by substations, large signal towers, etc. The optimization strategy is to adjust the frequency band or antenna tilt angle of the surrounding base stations.

[0082] If the clustered areas are isolated points or small clusters (such as an intersection) and the interference coefficient is abnormally high, they are identified as local voids, which may be caused by specific interference sources (such as illegal signal amplifiers). The optimization strategy is to add micro base stations near the point or adjust the local coverage.

[0083] In scenarios with strong electromagnetic interference, TCP layer retransmissions occur frequently due to interference. Using this as the core indicator for hole detection can easily lead to misjudgments, misidentifying interference-induced retransmissions as coverage holes. Therefore, in this scenario, the core indicator for evaluating service experience is switched from the general TCP retransmission rate to the application layer service success rate. This indicator is extracted from vehicle application logs, such as the percentage of successful reception of roadside perception information by an autonomous driving system within N consecutive seconds, or the stuttering rate of video playback in an in-vehicle entertainment system. This indicator directly reflects the user's actual experience and is more sensitive and accurate to the combined impact of underlying electromagnetic interference and coverage issues.

[0084] In the network operation and maintenance work order system, a hole repair effectiveness rating field (1-5 points, 1 point for no repair, 5 points for complete repair) is designed. Vehicle perception data within one week after repair is compared with that before repair, and a score is automatically generated (e.g., a 10dB improvement in RSRP results in a score of 4). Each quarter, hole cases with poor repair effectiveness (score ≤ 2 points) are included in the model training set to optimize the feature weights of the prediction model (e.g., if a region still experiences frequent false alarms after repair, the feature weight of the electromagnetic interference coefficient in that region is increased), allowing the model accuracy to gradually improve over time, reducing the target prediction error by 5%-8%.

[0085] In summary, this method can proactively and in real-time detect coverage holes, with positioning latency controlled within 10 seconds, achieving high-precision positioning at the sub-lane level. Through adaptive dynamic thresholds and a multi-model voting mechanism, the false alarm rate is reduced to below 5% and the false negative rate to below 3%. It also possesses interpretability, enabling the output of anomaly attribution through SHAP value analysis, providing clear directions for network optimization. Specialized adaptation algorithms are designed for extreme scenarios such as long tunnels, strong electromagnetic interference, and heavy rain / snowstorms to ensure full-scene coverage. Furthermore, closed-loop iteration is achieved based on the feedback of repair effect scores to continuously improve recognition accuracy.

[0086] This method boasts excellent deployment feasibility, with existing MR / DPI data acquisition systems deployed by operators, standard T-Boxes / RSUs in vehicles, and third-party meteorological data providing a solid data foundation. The optimized model can be deployed on existing cloud-based big data platforms, employing an edge computing and cloud aggregation architecture without requiring additional dedicated computing equipment. The work order system only needs to add a performance scoring field; model iteration can be automated via scripts, resulting in low maintenance costs and rapid deployment. This invention can be widely applied to network analysis, monitoring, and optimization in vertical scenarios such as connected vehicles and underground parking lots, providing operators with precise network optimization data and supporting the high-quality development of connected vehicle services.

[0087] Example 2 like Figure 2 As shown, this embodiment provides a vehicle network coverage hole identification system 200, including: Data acquisition module 201 is used to acquire multi-source heterogeneous data in the vehicle networking environment; Prediction module 202 is used to predict the network experience of each target vehicle within the coverage area of ​​the Internet of Vehicles at a target time point based on multi-source heterogeneous data, and obtain the expected network experience data of each target vehicle. Data acquisition module 203 is used to acquire actual network experience data of each target vehicle at the target time point; The anomaly identification module 204 is used to compare the expected network experience data of the target vehicle at the target time point with the actual network experience data of the target vehicle for each target vehicle, so as to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison result. The hole determination module 205 is used to aggregate all coverage anomalies within the same spatiotemporal range to obtain the coverage holes of the vehicle network.

[0088] Optionally, in an embodiment, the prediction module 202 includes: The data completion unit is used to perform spatiotemporal interpolation completion processing on multi-source heterogeneous data to obtain new multi-source heterogeneous data. The feature construction unit is used to construct features based on new multi-source heterogeneous data, and generate multi-dimensional fusion features for each target vehicle. The prediction unit is used to predict the network experience of each target vehicle at the target time point based on the multi-dimensional fusion features corresponding to the target vehicle, and obtain the expected network experience data of the target vehicle.

[0089] Optionally, in this embodiment, the data completion unit includes: The association graph generation sub-unit is used to construct a spatial association graph of each target vehicle with missing wireless network measurement data in multi-source heterogeneous data, with the target vehicle as the target node and the base stations and roadside units around the target vehicle as the surrounding nodes. The completion value calculation subunit is used to predict the wireless network measurement data completion value of each target vehicle with missing wireless network measurement data in multi-source heterogeneous data, based on the spatial association map of the target vehicle, through a pre-trained graph neural network based on the attention mechanism. The data completion subunit is used to fill in the multi-source heterogeneous data based on the completion values ​​of all wireless network measurement data to obtain new multi-source heterogeneous data.

[0090] Optionally, in the embodiments, the multi-dimensional fusion features include at least a signal stability index, an environmental interference coefficient, and a vehicle-road cooperative quality score; wherein, the signal stability index is data that quantifies the correlation between signal fluctuations and vehicle movement, the environmental interference coefficient is data that quantifies the impact of environmental factors on signal quality, and the vehicle-road cooperative quality score is data that quantifies the quality of the vehicle-road communication link.

[0091] Optionally, in an embodiment, the prediction unit includes: The first prediction subunit is used to input the multi-dimensional fusion features corresponding to the target vehicle into the pre-trained temporal prediction model to generate the first network experience expectation value. The topology graph construction sub-unit is used to construct a dynamic road network topology graph with the target vehicle and surrounding vehicles, base stations and roadside units as nodes; The second prediction subunit is used to generate a correction vector by propagating spatial information through a pre-trained dynamic graph neural network based on the dynamic road network topology map and the expected value of the first network experience. The correction subunit is used to correct the first network experience expectation value according to the correction vector to obtain the second network experience expectation value; The data generation subunit is used to obtain the network experience expectation data of the target vehicle based on the first network experience expectation value and the second network experience expectation value.

[0092] Optionally, in an embodiment, the anomaly detection module 204 includes: The first judgment unit is used to determine whether there is an anomaly in the location of the target vehicle based on the multi-dimensional fusion features corresponding to the target vehicle and through a pre-trained binary classification model, and obtain the first judgment result. The deviation calculation unit is used to calculate the first deviation between the first expected value of network experience and the actual data of network experience of the target vehicle, and to calculate the second deviation between the second expected value of network experience and the actual data of network experience of the target vehicle. The second judgment unit is used to compare the first deviation with the deviation threshold to obtain the second judgment result; The third judgment unit is used to compare the second deviation with the deviation threshold to obtain the third judgment result; The anomaly identification unit is used to make collaborative decisions through a voting mechanism based on the first judgment result, the second judgment result, and the third judgment result to determine whether the location of the target vehicle is a coverage anomaly point.

[0093] Optionally, in the embodiments, the deviation threshold is obtained in the following way: the vehicle network coverage area is divided into multiple spatiotemporal partitions based on the time and space dimensions; historical operation data in each spatiotemporal partition is obtained; for each spatiotemporal partition, the upper limit of the prediction error at a given confidence level is predicted by quantile regression based on the historical operation data in the spatiotemporal partition, which is used as the initial threshold of the spatiotemporal partition; the initial threshold corresponding to the spatiotemporal partition to which the target vehicle currently belongs is corrected based on the real-time environmental data in the multi-source heterogeneous data to obtain the deviation threshold.

[0094] Optionally, in an embodiment, the void determination module 205 includes: The anomaly point acquisition unit is used to acquire multiple covered anomaly points, each of which contains a geographic location and a timestamp. Clustering units are used to dynamically adjust the clustering radius based on the road type where the coverage anomalies are located. An improved DBSCAN algorithm is used to spatially cluster the coverage anomalies and generate candidate hole regions. The coverage hole generation unit is used to perform real-time streaming aggregation and merging of candidate hole regions according to a preset time window and spatial grid using a flow clustering algorithm, so as to generate vehicle network coverage holes and the geographical location boundary information of vehicle network coverage holes.

[0095] Optionally, in an embodiment, before the void determination module 205, the following method is further included: The first verification module is used to determine that the geographical location passes multi-vehicle verification if there are continuous abnormal coverage points from multiple vehicles of different models within a preset number of consecutive time windows for the same geographical location. The verification data acquisition module is used to acquire wireless network measurement data reported by the vehicle-mounted unit, communication quality data reported by the roadside unit and the corresponding vehicle-mounted unit, and regional signal quality data detected by the roadside unit for the geographical location verified by multiple vehicles. The second verification module is used to identify the coverage anomaly point corresponding to the geographical location as the coverage anomaly point for aggregation when the wireless network measurement data reported by the vehicle-mounted unit indicates that the signal strength is lower than the first preset threshold, and the communication quality data indicates communication abnormality or the regional signal quality data indicates that the signal strength is lower than the second preset threshold.

[0096] Optionally, in an embodiment, the system further includes: The light status acquisition module is used to acquire the vehicle light status of each target vehicle; The compensation module is used to perform environmental attenuation compensation on the expected network experience data based on the meteorological data in the multi-source heterogeneous data when the meteorological data in the multi-source heterogeneous data shows that the target weather exists, and the vehicle lights of more than a preset proportion of the target vehicles in the same area are in the fog light or low beam headlight state.

[0097] Optionally, in an embodiment, the prediction module 202 further includes: The tunnel scene adaptation unit is used to predict the network experience of the target vehicle when the target vehicle is in a tunnel, based on the vehicle location information in multi-source heterogeneous data and the pre-built signal attenuation reference curve, and obtain the expected network experience data of the target vehicle. The pre-built signal attenuation reference curve is a curve that is constructed based on the historical signal measurement data of the tunnel and is used to reflect the attenuation trend of the reference signal received power with the depth of the tunnel.

[0098] Optionally, in an embodiment, the system further includes: The heat map generation module is used to obtain the electromagnetic interference quantification indicators reported by multiple target vehicles and generate a spatial heat map of the interference source based on the electromagnetic interference quantification indicators. The void identification module is used to perform spatial clustering analysis based on the spatial heat map of the interference source in order to identify local covering voids.

[0099] In some embodiments, the vehicle network coverage hole identification system 200 of the present invention can be implemented in a combination of hardware and software. As an example, the vehicle network coverage hole identification system 200 of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the vehicle network coverage hole identification method of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0100] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0101] Example 3 like Figure 3 As shown, this embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the vehicle network coverage hole identification method as described in Embodiment 1.

[0102] In other words, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the vehicle network coverage hole identification method shown in any embodiment of the present invention by calling the computer program.

[0103] In one alternative embodiment, an electronic device is provided. Figure 3The illustrated electronic device 300 includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, via a bus 302. Optionally, the electronic device 300 may further include a transceiver 304, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 304 is not limited to one type, and the structure of the electronic device 300 does not constitute a limitation on the embodiments of the present invention.

[0104] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 301 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0105] Bus 302 may include a path for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 302 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus 302 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0106] The memory 303 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0107] The memory 303 is used to store application code (computer program) for executing the present invention, and its execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0108] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0109] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0110] Example 4 This embodiment provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the vehicle network coverage hole identification method of Embodiment 1.

[0111] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0112] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned method for identifying gaps in vehicle network coverage.

[0113] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer. In cases involving a remote computer, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0114] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0115] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0116] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0117] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0118] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0119] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0120] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for identifying a coverage hole of Internet of Vehicles, characterized in that, include: Acquire multi-source heterogeneous data in a connected vehicle environment; Based on the multi-source heterogeneous data, the network experience of each target vehicle within the coverage area of ​​the vehicle network at the target time point is predicted, and the expected network experience data of each target vehicle is obtained. Obtain actual network experience data for each target vehicle at the target time point; For each target vehicle, the expected network experience data of the target vehicle at the target time point is compared with the actual network experience data of the target vehicle, so as to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison result. By aggregating all coverage anomalies within the same spatiotemporal range, we obtain the coverage gaps in the vehicle-to-everything (V2X) network.

2. The method for identifying gaps in vehicle network coverage according to claim 1, characterized in that, The step of predicting the network experience of each target vehicle within the coverage area of ​​the vehicle network at a target time point based on the multi-source heterogeneous data, and obtaining the expected network experience data of each target vehicle, includes: Spatiotemporal interpolation completion processing is performed on the multi-source heterogeneous data to obtain new multi-source heterogeneous data; Based on the new multi-source heterogeneous data, feature construction is performed to generate multi-dimensional fusion features for each target vehicle. For each target vehicle, the network experience of the target vehicle at the target time point is predicted based on the multi-dimensional fusion features corresponding to the target vehicle, thereby obtaining the expected network experience data of the target vehicle.

3. The method for identifying gaps in vehicle network coverage according to claim 2, characterized in that, The process of performing spatiotemporal interpolation and completion on the multi-source heterogeneous data to obtain new multi-source heterogeneous data includes: For each target vehicle in the multi-source heterogeneous data where wireless network measurement data is missing, a spatial association map of the target vehicle is constructed, with the target vehicle as the target node and the base stations and roadside units around the target vehicle as the surrounding nodes. For each target vehicle with missing wireless network measurement data in the multi-source heterogeneous data, the wireless network measurement data completion value of the target vehicle is obtained by performing data prediction through a pre-trained graph neural network based on the spatial association map of the target vehicle. The multi-source heterogeneous data is filled in with the complete values ​​of all wireless network measurement data to obtain new multi-source heterogeneous data.

4. The method for identifying gaps in vehicle network coverage according to claim 2, characterized in that, The multidimensional fusion features include at least a signal stability index, an environmental interference coefficient, and a vehicle-road cooperative quality score; wherein, the signal stability index is data that quantifies the correlation between signal fluctuations and vehicle movement, the environmental interference coefficient is data that quantifies the impact of environmental factors on signal quality, and the vehicle-road cooperative quality score is data that quantifies the quality of the vehicle-road communication link.

5. The method for identifying gaps in vehicle network coverage according to claim 2, characterized in that, For each target vehicle, the network experience of the target vehicle at the target time point is predicted based on the multi-dimensional fusion features corresponding to the target vehicle, resulting in the expected network experience data of the target vehicle, including: The multidimensional fusion features corresponding to the target vehicle are input into a pre-trained temporal prediction model to generate the first network experience expectation value. A dynamic road network topology map is constructed using the target vehicle and surrounding vehicles, base stations, and roadside units as nodes; Based on the dynamic road network topology map and the first network experience expectation value, spatial information is propagated through a pre-trained dynamic graph neural network to generate a correction vector; The first network experience expectation value is corrected according to the correction vector to obtain the second network experience expectation value; Based on the first network experience expectation value and the second network experience expectation value, the network experience expectation data of the target vehicle is obtained.

6. The method for identifying gaps in vehicle network coverage according to claim 5, characterized in that, For each target vehicle, the expected network experience data and the actual network experience data of the target vehicle at the target time point are compared to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison result, including: Based on the multi-dimensional fusion features corresponding to the target vehicle, a pre-trained binary classification model is used to determine whether there is an anomaly in the location of the target vehicle, and a first judgment result is obtained. Calculate the first deviation between the first expected network experience value and the actual network experience data of the target vehicle, and calculate the second deviation between the second expected network experience value and the actual network experience data of the target vehicle; The first deviation is compared with the deviation threshold to obtain the second judgment result; The second deviation is compared with the deviation threshold to obtain a third judgment result; Based on the first judgment result, the second judgment result, and the third judgment result, a collaborative decision is made through a voting mechanism to determine whether the location of the target vehicle is a coverage anomaly.

7. The method for identifying gaps in vehicle network coverage according to claim 6, characterized in that, The deviation threshold is obtained in the following way: The coverage area of ​​the vehicle-to-everything (V2X) network is divided into multiple spatiotemporal partitions based on time and space dimensions. Obtain historical operational data for each spatiotemporal partition; For each spatiotemporal partition, based on the historical operational data within the spatiotemporal partition, the upper limit of the prediction error at a given confidence level is predicted using quantile regression, and this upper limit is used as the initial threshold for the spatiotemporal partition. The initial threshold corresponding to the current spatiotemporal partition of the target vehicle is corrected based on the real-time environmental data in the multi-source heterogeneous data to obtain the deviation threshold.

8. The method for identifying gaps in vehicle network coverage according to claim 1, characterized in that, The aggregation of all coverage anomalies within the same spatiotemporal range yields the vehicle network coverage gaps, including: Retrieve multiple coverage anomaly points, each containing a geographic location and a timestamp; The clustering radius is dynamically adjusted according to the road type where the coverage anomaly points are located, and the improved DBSCAN algorithm is used to perform spatial clustering of the coverage anomaly points to generate candidate hole regions. A flow clustering algorithm is used to perform real-time flow aggregation and merging of the candidate hole regions according to a preset time window and spatial grid, generating vehicle network coverage holes and the geographical location boundary information of the vehicle network coverage holes.

9. The method for identifying gaps in vehicle network coverage according to claim 1, characterized in that, Before aggregating all coverage anomalies within the same spatiotemporal range to obtain the vehicle network coverage gaps, the following steps are also included: For the same geographical location, if there are continuous coverage anomalies from multiple vehicles of different models within a preset number of consecutive time windows, then the geographical location is determined to have passed multi-vehicle verification. For geographical locations verified by multiple vehicles, obtain wireless network measurement data reported by the vehicle-mounted unit, communication quality data reported by the roadside unit and the corresponding vehicle-mounted unit, and regional signal quality data detected by the roadside unit. When the wireless network measurement data reported by the vehicle-mounted system indicates that the signal strength is lower than a first preset threshold, and the communication quality data indicates a communication anomaly or the regional signal quality data indicates that the signal strength is lower than a second preset threshold, the coverage anomaly point corresponding to the geographical location is identified as a coverage anomaly point for aggregation.

10. The method for identifying gaps in vehicle network coverage according to claim 1, characterized in that, Also includes: Obtain the vehicle lighting status of each target vehicle; When the meteorological data in the multi-source heterogeneous data shows that there is target weather, and the vehicle lights of more than a preset proportion of target vehicles in the same area are in fog light or low beam mode, environmental attenuation compensation is performed on the network experience expectation data based on the meteorological data in the multi-source heterogeneous data.

11. The method for identifying gaps in vehicle network coverage according to claim 1, characterized in that, The step of predicting the network experience of each target vehicle within the coverage area of ​​the vehicle network at a target time point based on the multi-source heterogeneous data, and obtaining the expected network experience data of each target vehicle, further includes: When the target vehicle is in a tunnel, the network experience of the target vehicle is predicted based on the vehicle location information in the multi-source heterogeneous data and the pre-constructed signal attenuation reference curve, so as to obtain the expected network experience data of the target vehicle. The pre-constructed signal attenuation reference curve is a curve constructed based on historical signal measurement data of the tunnel, used to reflect the attenuation trend of the reference signal received power as the tunnel deepens.

12. The method for identifying gaps in vehicle network coverage according to claim 1, characterized in that, Also includes: The electromagnetic interference quantification index reported by multiple target vehicles is obtained, and a spatial heat map of the interference source is generated based on the electromagnetic interference quantification index. Spatial clustering analysis is performed based on the spatial heat map of the interference source to identify local coverage voids.

13. A vehicle network coverage hole identification system, characterized in that, include: The data acquisition module is used to acquire multi-source heterogeneous data in the vehicle-to-everything (V2X) environment; The prediction module is used to predict the network experience of each target vehicle within the coverage area of ​​the Internet of Vehicles at a target time point based on the multi-source heterogeneous data, and obtain the expected network experience data of each target vehicle. The data acquisition module is used to acquire the actual network experience data of each target vehicle at the target time point; An anomaly detection module is used to compare the expected network experience data of the target vehicle and the actual network experience data of the target vehicle at the target time point for each target vehicle, so as to determine whether the location of the target vehicle is a coverage anomaly point based on the comparison result. The hole determination module is used to aggregate all coverage anomalies within the same spatiotemporal range to obtain the coverage holes of the vehicle network.

14. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the vehicle network coverage hole identification method as described in any one of claims 1 to 12.

15. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the vehicle network coverage hole identification method according to any one of claims 1 to 12.

Images